Strange lumps of flesh float in space, with limb-like protrusions and peachy flesh snaking over a darkened backdrop. Many of the forms recall Salvador Dali’s barely-identifiable melting creatures. One in particular looks like a hallucinatory dream of the Venus de Milo. Another looks like an impressionist portrait of a woman lounging in a divan. However, these surreal images were actually created by a neural network, trained on nude portraits (mostly women). The neural network, trained on human culture, reflects art history’s fixation on the female form back at us.
Here are some AI generated nude portraits I've been working on????
Usually the machine just paints people as blobs of flesh with tendrils and limbs randomly growing out – I think it's really surreal. I wonder if that's how machines see us… pic.twitter.com/tYgzCHGfse
— Robbie Barrat (@DrBeef_) March 27, 2018
The images were produced by AI researcher Robbie Barrat, who posts his experiments on Twitter (“please don’t tell my manager,” he notes).
Barrat also trains GANs, or generative adversarial networks, on classical landscape paintings, tweeting that the following image took two weeks to complete:
Exploring the latent space of AI-generated landscape paintings today instead of doing work.
Please do not tell my manager. pic.twitter.com/vNVsPMnMkz
— Robbie Barrat (@DrBeef_) March 19, 2018
On Twitter, Barrat cites the artist Sol LeWitt’s work as an analog to this algorithmic art, because LeWitt would create sets of rules for his works that others could then interpret as they wanted. There’s one significant distinction though. “I think the really interesting difference between actual computers and people as computers is that people can interpret rules, and not just follow a strict set of programmed rules perfectly,” he writes on Twitter. “AI created art is different than any other digital generative art for this reason.”
Pulsating AI generated nude portrait ???? pic.twitter.com/PrnNSi53jb
— Robbie Barrat (@DrBeef_) March 29, 2018
Though his research is on the cutting edge of machine learning, Barrat himself recently graduated from high school in West Virginia–after a stint at Nvidia, he’s currently working in a research lab at Stanford.